The Onset of Molecule-Spanning Dynamics in Heat Shock Protein Hsp90

Abstract

Protein dynamics have been investigated on a wide range of time scales. Nano- and picosecond dynamics have been assigned to local fluctuations, while slower dynamics have been attributed to larger conformational changes. However, it is largely unknown how fast (local) fluctuations can lead to slow global (allosteric) changes. Here, fast molecule-spanning dynamics on the 100 to 200 ns time scale in the heat shock protein 90 (Hsp90) are shown. Global real-space movements are assigned to dynamic modes on this time scale, which is possible by a combination of single-molecule fluorescence, quasi-elastic neutron scattering and all-atom molecular dynamics (MD) simulations. The time scale of these dynamic modes depends on the conformational state of the Hsp90 dimer. In addition, the dynamic modes are affected to various degrees by Sba1, a co-chaperone of Hsp90, depending on the location within Hsp90, which is in very good agreement with MD simulations. Altogether, this data is best described by fast molecule-spanning dynamics, which precede larger conformational changes in Hsp90 and might be the molecular basis for allostery. This integrative approach provides comprehensive insights into molecule-spanning dynamics on the nanosecond time scale for a multi-domain protein.

Keywords: heat shock protein 90; molecular dynamics simulations; neutron scattering; protein dynamics; single molecule fluorescence.

Figure 1

Comprehensive picture of Hsp90 dynamics. Here, we complete this picture by accessing the 150 ns time scale, which we show to unravel molecule‐spanning dynamics. Disentanglement of rotational and translational dynamics (red and blue frames, respectively) is feasible by combining nanosecond fluorescence correlation spectroscopy (nsFCS), time‐resolved anisotropy (TRA), neutron spin echo spectroscopy (NSE), neutron backscattering spectroscopy (NBS) and molecular dynamics (MD) simulations. The time scales (from experiments) of the respective dynamics are given in the boxes. TRA is only sensitive to rotations, while all other methods are sensitive to rotation and translation.

Figure 2

Substate‐specific nsFCS analysis of the Hsp90 FRET pair 452‐452 labeled with Atto550 and Atto647N with AMPPNP. The results of all six independent measurements are given in Table S1, Supporting Information. Exemplary data from one measurement is shown here. a) Scheme of a nsFCS experiment. b) Substate‐selection based on the FRET efficiency E. Selecting single‐molecule events with E = 0.4–0.8 (light blue part) enables specific analysis of the closed A conformation of Hsp90. c) Logarithmic substate‐specific polarized‐FCS curves from seconds to picoseconds correlating the parallel and the perpendicular detection signal and vice versa. Fit models include diffusion, bunching and antibunching components, for the autocorrelations (D×D and A×A) additionally a triplet kinetics component was used. d) Linear substate‐specific nsFCS curves. From a global analysis of all three channel correlations (D×D, A×D, and A×A) we obtain a correlation time of (149  ±  7) ns. For the complete nsFCS analysis see Table S1, Supporting Information. Errors were determined as standard errors of the mean for the six independent measurements. e) Subtraction of the two not normalized nsFCS branches shows the influence of polarization and therefore rotation on the observed dynamics. Shown in black are double exponential decays with relaxation times of 3 and 60 ns. Both are much shorter than the 149 ns dynamics observed in (d).

Figure 3

Substate‐specific time‐resolved single‐molecule anisotropy of the Hsp90 FRET pair 452‐452 with AMPPNP. a) The transfer efficiency histogram reveals three conformational states of Hsp90: the open state (E ≈ 0.17, highlighted blue), closed state A (E ≈ 0.6, highlighted green), and closed state B (E ≈ 0.9, highlighted red). For the following graphs, donor‐only and acceptor‐only molecules were excluded by selecting exclusively single‐molecule events via dual channel burst search algorithm for further analysis. b) Schematic view of an anisotropy experiment. Fluorescence depolarization depends on the lifetime and rotation of the excited dye dipole and is measured by polarization‐sensitive detection. c) Acceptor anisotropy decay upon direct acceptor excitation at Hsp90 FRET pair 452‐452 for the three different populations. In black a cone‐in‐cone model fit is shown. r ₀ = 0.4 and global parameters ρ_dye and ρ_global result in a rotational decay time of (52  ±  32) ns. Fit results and standard errors for all sub‐populations are shown in Table S2, Supporting Information. Figure S5 and Table S6, Supporting Information, show the results for other fit functions that give comparable values.

Figure 4

MD simulations confirm rotational correlation and provide a link between nsFCS and dye accessible volumes. a) Autocorrelation of Hsp90's first principal axis of the moment of inertia orientation reveals a rotational correlation time of (81 ± 1) ns. Shaded area depicts the standard error of the mean for five statistically independent simulations. b) A representative MD snapshot of the Hsp90 dimer visualizes the investigated accessible dye volumes (colored spheres). Hsp90 chains A and B are colored gray and green, respectively. c) Autocorrelation functions of accessible dye volumes at Hsp90 positions 61 and 452 at chain A or B, respectively. Unconstrained bi‐exponential fits reveal correlation times on the 100 ns time scale, which might be the cause for fluctuations in fluorescence intensity resulting from structural dynamics (see Figure S6, Supporting Information).

Figure 5

Neutron spin echo spectroscopy shows internal motions beyond the translation and rotation of the entire molecule. a) Fits of the intermediate scattering functions for different scattering vectors q (color coded). The fits (solid lines) were performed using Equation (2) for τ < 30 ns and for I(q, τ) > 0.3. They were extrapolated with dotted lines to outline the expected deviation from the single exponential suggesting the presence of internal motions. b) The resulting experimental diffusion function D(q) evidences a first shoulder around the scattering vector q ≈ 0.07 Å⁻¹ and a peak around q ≈ 0.13 Å⁻¹. While the shoulder can be described by rotational diffusion of the entire protein based on a rigid‐body modeling, the peak originates from an internal degree of freedom. c) Visualization of the first and most relevant principal component of Hsp90: a bending motion with a slight twist spans over the entire molecule (left). Calculated amplitude functions A(q) based on Equation (3) signatures of the ten PCA eigenvectors obtained from five MD runs with independent start parameters (see main text for details). The peak around q ≈ 0.13 Å⁻¹ indicates that Hsp90 performs concerted internal motions on times scales of 100 ns (right), consistent with nsFCS.

Figure 6

Modulation of Hsp90's nanosecond dynamics. Substate‐specific nsFCS analysis of the same construct as in Figure 2, namely the Hsp90 FRET pair 452‐452. a) Measurement in the presence of AMPPNP and Sba1 to further populate the closed state A and investigate the effect of Sba1. The bunching component is around (151  ±  14) ns, not significantly different from the value in absence of Sba1. b) Measurement in the absence of AMPPNP to populate the open state of Hsp90 and the analysis of this open state (E = 0.1–0.4). The bunching component is (187  ±  10) ns and therefore significantly different from the closed state A.

Figure 7

Molecule‐spanning dynamics on the 150 ns time scale might be precursors of large conformational changes between two equilibrium states (marked with red stars). They probe higher regions of the free energy landscape and likely guide the protein toward the point of no return (PNR) to complete, for example, an allosteric signaling path. Only one exemplary path, which might result from the highly diffusive motions, is depicted here.